ANN_CNN.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "JTgA4yl3sDkH"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "QhvucQi8rjAp"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import os\n",
        "import pandas as pd\n",
        "from keras.utils import np_utils\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense\n",
        "from sklearn.model_selection import train_test_split\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers.convolutional import MaxPooling2D\n",
        "from keras.layers import Dropout\n",
        "from keras.constraints import maxnorm\n",
        "from keras.layers import activation"
      ],
      "metadata": {
        "id": "lIBjEzJ-tm8M"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "mK2rlMqqrlN-"
      },
      "outputs": [],
      "source": [
        "file = \"/content/drive/MyDrive/cifar-10-batches-py/\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "lMFXqelTsAgz"
      },
      "outputs": [],
      "source": [
        "def unpickle(file):\n",
        "    import pickle\n",
        "    with open(file, 'rb') as fo:\n",
        "        dict = pickle.load(fo, encoding='bytes')\n",
        "    return dict"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "jiOEeNDZsFNc"
      },
      "outputs": [],
      "source": [
        "def load_batch_file(batch_filename):\n",
        "    filepath = os.path.join(file, batch_filename)\n",
        "    unpickled = unpickle(filepath)\n",
        "    return unpickled"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "MNCohiz4sG4y"
      },
      "outputs": [],
      "source": [
        "train_batch_1 = load_batch_file('data_batch_1')\n",
        "train_batch_2 = load_batch_file('data_batch_2')\n",
        "train_batch_3 = load_batch_file('data_batch_3')\n",
        "train_batch_4 = load_batch_file('data_batch_4')\n",
        "train_batch_5 = load_batch_file('data_batch_5')\n",
        "test_batch = load_batch_file('test_batch')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m9bJny8VsKy4"
      },
      "source": [
        "Concatinating all test batch data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "uPKVjPC1sG_x"
      },
      "outputs": [],
      "source": [
        "num_classes = 10\n",
        "Total_train_x = np.concatenate([train_batch_1[b'data'], train_batch_2[b'data'], train_batch_3[b'data'], train_batch_4[b'data'], train_batch_5[b'data']])\n",
        "Total_train_x = Total_train_x.astype('float32') # this is necessary for the division below\n",
        "Total_train_x /= 255\n",
        "Total_train_y = np.concatenate([np_utils.to_categorical(labels, num_classes) for labels in [train_batch_1[b'labels'], train_batch_2[b'labels'], train_batch_3[b'labels'], train_batch_4[b'labels'], train_batch_5[b'labels']]])"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "Total_train_y.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cXnYk22Ux56T",
        "outputId": "42a423d3-0a54-4015-dcab-701288c78f83"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(50000, 10)"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_train,x_val,y_train,y_val=train_test_split(Total_train_x,Total_train_y,test_size=0.2)\n"
      ],
      "metadata": {
        "id": "3mjd3phTpKZL"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "Total_train_x.shape\n",
        "Total_train_y.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tdE82T6xrA8p",
        "outputId": "965e8777-6775-42b3-c005-0af0283053e3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(50000, 10)"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_train.shape\n",
        "y_train.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1GWr5OA9rHMv",
        "outputId": "9963cae3-a78b-47b5-a5db-ea04e5f8de04"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(40000, 10)"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_val.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "naMni6n81eis",
        "outputId": "3af52d5a-272d-491b-a43d-6593cdc1ff1b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(10000, 3072)"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "y_val.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EqvwDAZArOTI",
        "outputId": "698e3e8c-ca3b-47c1-bdcf-c277389bd1b4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(10000, 10)"
            ]
          },
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X0PvZDeisRD5"
      },
      "source": [
        "**Test batch data**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "0E_W0YLusKdE"
      },
      "outputs": [],
      "source": [
        "test_x = test_batch[b'data'].astype('float32') / 255\n",
        "test_y = np_utils.to_categorical(test_batch[b'labels'], num_classes)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "id": "OA4xmG0RsU-e"
      },
      "outputs": [],
      "source": [
        "img_rows = img_cols = 32\n",
        "channels = 3"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pTq7iLyYMQ8r"
      },
      "source": [
        "## **Neural Network training**"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "8QqVZ9RpxWJd"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Technique#1** (Concatanating all data and then splitting to 20-80 for validation & training data)"
      ],
      "metadata": {
        "id": "PA-JVgheRKUa"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "simple_model = Sequential()\n",
        "simple_model.add(Dense(10_000, input_shape=(img_rows*img_cols*channels,), activation='relu'))\n",
        "simple_model.add(Dense(1_000, activation='relu'))\n",
        "simple_model.add(Dense(500, activation='relu'))\n",
        "simple_model.add(Dense(250, activation='relu'))\n",
        "simple_model.add(Dense(100, activation='relu'))\n",
        "simple_model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "simple_model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "simple_model_history = simple_model.fit(x_train, y_train, batch_size=50, epochs=15, validation_data=(x_val, y_val))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-Q6pRM04DpC2",
        "outputId": "256637d3-950f-4a7e-de5a-a0b3277f6ade"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/15\n",
            "800/800 [==============================] - 244s 304ms/step - loss: 1.9131 - accuracy: 0.3073 - val_loss: 1.7479 - val_accuracy: 0.3695\n",
            "Epoch 2/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.7040 - accuracy: 0.3921 - val_loss: 1.6306 - val_accuracy: 0.4249\n",
            "Epoch 3/15\n",
            "800/800 [==============================] - 239s 299ms/step - loss: 1.6111 - accuracy: 0.4284 - val_loss: 1.5739 - val_accuracy: 0.4307\n",
            "Epoch 4/15\n",
            "800/800 [==============================] - 239s 299ms/step - loss: 1.5441 - accuracy: 0.4507 - val_loss: 1.5304 - val_accuracy: 0.4500\n",
            "Epoch 5/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.4895 - accuracy: 0.4710 - val_loss: 1.4963 - val_accuracy: 0.4651\n",
            "Epoch 6/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.4393 - accuracy: 0.4909 - val_loss: 1.4529 - val_accuracy: 0.4881\n",
            "Epoch 7/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.3988 - accuracy: 0.5035 - val_loss: 1.4426 - val_accuracy: 0.4790\n",
            "Epoch 8/15\n",
            "800/800 [==============================] - 239s 299ms/step - loss: 1.3608 - accuracy: 0.5176 - val_loss: 1.3986 - val_accuracy: 0.5026\n",
            "Epoch 9/15\n",
            "800/800 [==============================] - 239s 299ms/step - loss: 1.3221 - accuracy: 0.5311 - val_loss: 1.3984 - val_accuracy: 0.4967\n",
            "Epoch 10/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.2859 - accuracy: 0.5460 - val_loss: 1.3980 - val_accuracy: 0.5013\n",
            "Epoch 11/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.2516 - accuracy: 0.5583 - val_loss: 1.4241 - val_accuracy: 0.4934\n",
            "Epoch 12/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.2135 - accuracy: 0.5699 - val_loss: 1.3971 - val_accuracy: 0.5040\n",
            "Epoch 13/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.1846 - accuracy: 0.5803 - val_loss: 1.4161 - val_accuracy: 0.5034\n",
            "Epoch 14/15\n",
            "800/800 [==============================] - 256s 320ms/step - loss: 1.1484 - accuracy: 0.5929 - val_loss: 1.3810 - val_accuracy: 0.5115\n",
            "Epoch 15/15\n",
            "800/800 [==============================] - 246s 307ms/step - loss: 1.1159 - accuracy: 0.6043 - val_loss: 1.3461 - val_accuracy: 0.5216\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "def plot_history(history, title):\n",
        "    plt.figure(figsize=(10,3))\n",
        "    # Plot training & validation accuracy values\n",
        "    plt.subplot(121)\n",
        "    plt.plot(history.history['accuracy'])\n",
        "    plt.plot(history.history['val_accuracy'])\n",
        "    plt.title('Model accuracy')\n",
        "    plt.ylabel('Accuracy')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "\n",
        "    # Plot training & validation loss values\n",
        "    plt.subplot(122)\n",
        "    plt.plot(history.history['loss'])\n",
        "    plt.plot(history.history['val_loss'])\n",
        "    plt.title('Model loss')\n",
        "    plt.ylabel('Loss')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "    plt.show()"
      ],
      "metadata": {
        "id": "_VC7X0sBpEjg"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plot_history(simple_model_history, 'Simple NN with batch size=50 epochs=15 ')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "id": "glNybEmJrH_C",
        "outputId": "f85fb1f2-55d9-4008-f548-6e3d483a8594"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 720x216 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Evaluating Accuracy through prediction on test data**"
      ],
      "metadata": {
        "id": "9Ag3FWtvrqN2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "simple_model.evaluate(test_x,test_y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rMyTIbGVrxOG",
        "outputId": "0054c5d8-eb75-446f-8128-335ad94ee0cf"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 29s 92ms/step - loss: 1.3446 - accuracy: 0.5162\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.3446089029312134, 0.5162000060081482]"
            ]
          },
          "metadata": {},
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Time Taken by model to train = **3622 sec** = **60 mins**\n",
        "## **241 sec** => avg time of each epoch\n",
        "\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "oQ6-4v5XsnP2"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**TECHNIQUE # 2 (Cross validation)**"
      ],
      "metadata": {
        "id": "oJmmqeK9dkcF"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "num_classes = 10\n",
        "Total_train_x = np.concatenate( [train_batch_2[b'data'], train_batch_3[b'data'], train_batch_4[b'data'], train_batch_5[b'data']])\n",
        "Total_train_x = Total_train_x.astype('float32') # this is necessary for the division below\n",
        "Total_train_x /= 255\n",
        "Total_train_y = np.concatenate([np_utils.to_categorical(labels, num_classes) for labels in [ train_batch_2[b'labels'], train_batch_3[b'labels'], train_batch_4[b'labels'], train_batch_5[b'labels']]])"
      ],
      "metadata": {
        "id": "p-vz40uGl1Gj"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_val = np.concatenate([train_batch_1[b'data']])\n",
        "x_val = x_val.astype('float32')\n",
        "x_val /= 255 \n",
        "x_val.shape\n",
        "\n",
        "y_val = np_utils.to_categorical(train_batch_1[b'labels'], num_classes)\n",
        "# y_val"
      ],
      "metadata": {
        "id": "qgXVkuuldUAD"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "test_x = test_batch[b'data'].astype('float32') / 255\n",
        "test_y = np_utils.to_categorical(test_batch[b'labels'], num_classes)"
      ],
      "metadata": {
        "id": "0eplzzoasFpl"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "img_rows = img_cols = 32\n",
        "channels = 3"
      ],
      "metadata": {
        "id": "uV9u3AH7sFmr"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "Total_train_x.shape\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TuRQRq2Hw926",
        "outputId": "19fd9d7f-ca01-49ab-bf53-1ea03e6cd1b2"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(40000, 3072)"
            ]
          },
          "metadata": {},
          "execution_count": 122
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "Total_train_y.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sJpw72z3xED3",
        "outputId": "7fbb3d0c-65dd-4bae-8aa8-b5e0e9df3c5c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(40000, 10)"
            ]
          },
          "metadata": {},
          "execution_count": 123
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "simple_model = Sequential()\n",
        "simple_model.add(Dense(10_000, input_shape=(img_rows*img_cols*channels,), activation='relu'))\n",
        "simple_model.add(Dense(1_000, activation='relu'))\n",
        "simple_model.add(Dense(500, activation='relu'))\n",
        "simple_model.add(Dense(250, activation='relu'))\n",
        "simple_model.add(Dense(100, activation='relu'))\n",
        "simple_model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "simple_model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "simple_model_history = simple_model.fit(Total_train_x, Total_train_y, batch_size=50, epochs=15, validation_data=(x_val, y_val))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DAI9SqPTtOgA",
        "outputId": "cfd814a8-c87a-4d32-bbde-c0b28032e30a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/15\n",
            "800/800 [==============================] - 279s 305ms/step - loss: 1.9211 - accuracy: 0.3058 - val_loss: 1.7430 - val_accuracy: 0.3700\n",
            "Epoch 2/15\n",
            "800/800 [==============================] - 244s 305ms/step - loss: 1.7016 - accuracy: 0.3945 - val_loss: 1.6564 - val_accuracy: 0.4041\n",
            "Epoch 3/15\n",
            "800/800 [==============================] - 243s 304ms/step - loss: 1.6115 - accuracy: 0.4266 - val_loss: 1.6121 - val_accuracy: 0.4164\n",
            "Epoch 4/15\n",
            "800/800 [==============================] - 244s 305ms/step - loss: 1.5473 - accuracy: 0.4516 - val_loss: 1.5241 - val_accuracy: 0.4546\n",
            "Epoch 5/15\n",
            "800/800 [==============================] - 246s 308ms/step - loss: 1.4936 - accuracy: 0.4678 - val_loss: 1.4884 - val_accuracy: 0.4666\n",
            "Epoch 6/15\n",
            "800/800 [==============================] - 247s 309ms/step - loss: 1.4440 - accuracy: 0.4878 - val_loss: 1.4605 - val_accuracy: 0.4795\n",
            "Epoch 7/15\n",
            "800/800 [==============================] - 246s 307ms/step - loss: 1.4010 - accuracy: 0.5040 - val_loss: 1.4481 - val_accuracy: 0.4793\n",
            "Epoch 8/15\n",
            "800/800 [==============================] - 242s 303ms/step - loss: 1.3623 - accuracy: 0.5167 - val_loss: 1.4139 - val_accuracy: 0.4952\n",
            "Epoch 9/15\n",
            "800/800 [==============================] - 241s 301ms/step - loss: 1.3262 - accuracy: 0.5302 - val_loss: 1.4254 - val_accuracy: 0.4909\n",
            "Epoch 10/15\n",
            "800/800 [==============================] - 239s 299ms/step - loss: 1.2893 - accuracy: 0.5408 - val_loss: 1.4490 - val_accuracy: 0.4889\n",
            "Epoch 11/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.2498 - accuracy: 0.5561 - val_loss: 1.4151 - val_accuracy: 0.5018\n",
            "Epoch 12/15\n",
            "800/800 [==============================] - 257s 321ms/step - loss: 1.2195 - accuracy: 0.5682 - val_loss: 1.3696 - val_accuracy: 0.5203\n",
            "Epoch 13/15\n",
            "800/800 [==============================] - 240s 301ms/step - loss: 1.1838 - accuracy: 0.5808 - val_loss: 1.3707 - val_accuracy: 0.5100\n",
            "Epoch 14/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.1521 - accuracy: 0.5949 - val_loss: 1.3093 - val_accuracy: 0.5373\n",
            "Epoch 15/15\n",
            "800/800 [==============================] - 240s 300ms/step - loss: 1.1197 - accuracy: 0.6050 - val_loss: 1.3981 - val_accuracy: 0.5154\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "simple_model.predict(test_x)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QZys-Ai7tayO",
        "outputId": "2edd0abf-3344-48a3-e420-6f08500e5acc"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 29s 91ms/step\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[7.4106559e-02, 4.1388761e-02, 7.4662536e-02, ..., 2.7139890e-03,\n",
              "        1.2088811e-01, 1.2894733e-02],\n",
              "       [1.2999491e-02, 6.5660290e-02, 2.8404983e-04, ..., 1.2066223e-04,\n",
              "        2.8086641e-01, 6.3965881e-01],\n",
              "       [2.8915805e-01, 8.5316204e-02, 1.7148637e-03, ..., 4.9742446e-03,\n",
              "        4.4786483e-01, 1.6708985e-01],\n",
              "       ...,\n",
              "       [2.9521899e-03, 8.2105675e-05, 1.7776006e-01, ..., 5.8499236e-02,\n",
              "        2.1794096e-02, 3.5011028e-03],\n",
              "       [1.0118003e-01, 3.4500111e-02, 1.7061195e-01, ..., 2.1921799e-02,\n",
              "        4.1962188e-02, 3.5615548e-02],\n",
              "       [5.7914287e-02, 6.6417181e-03, 5.2016180e-02, ..., 5.5094504e-01,\n",
              "        8.4964018e-03, 3.0101856e-02]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 126
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "simple_model.evaluate(test_x,test_y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "M60XHZx8MrKY",
        "outputId": "e1c473dd-1fb6-4605-9934-2cc5f91bb7e6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 27s 84ms/step - loss: 1.3995 - accuracy: 0.5118\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.3995152711868286, 0.5117999911308289]"
            ]
          },
          "metadata": {},
          "execution_count": 127
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Convolution Neural Network**"
      ],
      "metadata": {
        "id": "pMyyb39fsFQ4"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "ez96yX80NyDY"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_x_reshaped = x_train.reshape(len(x_train), img_rows, img_cols, channels)\n",
        "val_x_reshaped = x_val.reshape(len(x_val), img_rows, img_cols, channels)\n",
        "test_x_reshaped = test_x.reshape(len(test_x), img_rows, img_cols, channels)"
      ],
      "metadata": {
        "id": "TSkNDUPysb6e"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers import Conv2D, Flatten\n",
        "simple_cnn_model = Sequential()\n",
        "#---------------------------------------------------------------------------------------------------------------------s\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), input_shape=(img_rows,img_cols,channels), activation='relu')) \n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "\n",
        "simple_cnn_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "simple_cnn_model.add(Dropout(0.25))\n",
        "simple_cnn_model.add(Flatten())\n",
        " \n",
        "simple_cnn_model.add(Dense(128, activation='relu'))\n",
        "simple_cnn_model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "\n",
        "simple_cnn_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "simple_cnn_model_history = simple_cnn_model.fit(train_x_reshaped, y_train, batch_size=10, epochs=35, validation_data=(val_x_reshaped, y_val))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kMY4JaVosgrV",
        "outputId": "062eaa8e-4ac6-4c20-cdb6-2909ace2e734"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/35\n",
            "4000/4000 [==============================] - 365s 91ms/step - loss: 1.6175 - accuracy: 0.4126 - val_loss: 1.3696 - val_accuracy: 0.5122\n",
            "Epoch 2/35\n",
            "4000/4000 [==============================] - 350s 88ms/step - loss: 1.3318 - accuracy: 0.5267 - val_loss: 1.3012 - val_accuracy: 0.5397\n",
            "Epoch 3/35\n",
            "4000/4000 [==============================] - 349s 87ms/step - loss: 1.1994 - accuracy: 0.5745 - val_loss: 1.2783 - val_accuracy: 0.5509\n",
            "Epoch 4/35\n",
            "4000/4000 [==============================] - 349s 87ms/step - loss: 1.1013 - accuracy: 0.6096 - val_loss: 1.1870 - val_accuracy: 0.5830\n",
            "Epoch 5/35\n",
            "4000/4000 [==============================] - 348s 87ms/step - loss: 1.0141 - accuracy: 0.6416 - val_loss: 1.1862 - val_accuracy: 0.5863\n",
            "Epoch 6/35\n",
            "4000/4000 [==============================] - 348s 87ms/step - loss: 0.9469 - accuracy: 0.6627 - val_loss: 1.2004 - val_accuracy: 0.5845\n",
            "Epoch 7/35\n",
            "4000/4000 [==============================] - 348s 87ms/step - loss: 0.8832 - accuracy: 0.6859 - val_loss: 1.2355 - val_accuracy: 0.5848\n",
            "Epoch 8/35\n",
            "4000/4000 [==============================] - 349s 87ms/step - loss: 0.8280 - accuracy: 0.7080 - val_loss: 1.1798 - val_accuracy: 0.6019\n",
            "Epoch 9/35\n",
            "4000/4000 [==============================] - 351s 88ms/step - loss: 0.7711 - accuracy: 0.7238 - val_loss: 1.2487 - val_accuracy: 0.5887\n",
            "Epoch 10/35\n",
            "4000/4000 [==============================] - 353s 88ms/step - loss: 0.7302 - accuracy: 0.7380 - val_loss: 1.3033 - val_accuracy: 0.5868\n",
            "Epoch 11/35\n",
            "4000/4000 [==============================] - 356s 89ms/step - loss: 0.6962 - accuracy: 0.7512 - val_loss: 1.3045 - val_accuracy: 0.5833\n",
            "Epoch 12/35\n",
            "4000/4000 [==============================] - 359s 90ms/step - loss: 0.6625 - accuracy: 0.7649 - val_loss: 1.3323 - val_accuracy: 0.5871\n",
            "Epoch 13/35\n",
            "4000/4000 [==============================] - 364s 91ms/step - loss: 0.6239 - accuracy: 0.7759 - val_loss: 1.4249 - val_accuracy: 0.5797\n",
            "Epoch 14/35\n",
            "4000/4000 [==============================] - 367s 92ms/step - loss: 0.5946 - accuracy: 0.7888 - val_loss: 1.4413 - val_accuracy: 0.5862\n",
            "Epoch 15/35\n",
            "4000/4000 [==============================] - 366s 92ms/step - loss: 0.5713 - accuracy: 0.7966 - val_loss: 1.4527 - val_accuracy: 0.5801\n",
            "Epoch 16/35\n",
            "4000/4000 [==============================] - 372s 93ms/step - loss: 0.5600 - accuracy: 0.8007 - val_loss: 1.4632 - val_accuracy: 0.5790\n",
            "Epoch 17/35\n",
            "4000/4000 [==============================] - 374s 94ms/step - loss: 0.5283 - accuracy: 0.8100 - val_loss: 1.5532 - val_accuracy: 0.5875\n",
            "Epoch 18/35\n",
            "4000/4000 [==============================] - 381s 95ms/step - loss: 0.5067 - accuracy: 0.8183 - val_loss: 1.5520 - val_accuracy: 0.5699\n",
            "Epoch 19/35\n",
            "4000/4000 [==============================] - 379s 95ms/step - loss: 0.4933 - accuracy: 0.8246 - val_loss: 1.5912 - val_accuracy: 0.5802\n",
            "Epoch 20/35\n",
            "4000/4000 [==============================] - 379s 95ms/step - loss: 0.4869 - accuracy: 0.8249 - val_loss: 1.6078 - val_accuracy: 0.5811\n",
            "Epoch 21/35\n",
            "4000/4000 [==============================] - 379s 95ms/step - loss: 0.4712 - accuracy: 0.8321 - val_loss: 1.6357 - val_accuracy: 0.5754\n",
            "Epoch 22/35\n",
            "4000/4000 [==============================] - 382s 95ms/step - loss: 0.4613 - accuracy: 0.8360 - val_loss: 1.6599 - val_accuracy: 0.5771\n",
            "Epoch 23/35\n",
            "4000/4000 [==============================] - 379s 95ms/step - loss: 0.4498 - accuracy: 0.8395 - val_loss: 1.6805 - val_accuracy: 0.5771\n",
            "Epoch 24/35\n",
            "4000/4000 [==============================] - 387s 97ms/step - loss: 0.4375 - accuracy: 0.8448 - val_loss: 1.6843 - val_accuracy: 0.5789\n",
            "Epoch 25/35\n",
            "4000/4000 [==============================] - 393s 98ms/step - loss: 0.4269 - accuracy: 0.8498 - val_loss: 1.6358 - val_accuracy: 0.5836\n",
            "Epoch 26/35\n",
            "4000/4000 [==============================] - 388s 97ms/step - loss: 0.4241 - accuracy: 0.8501 - val_loss: 1.6793 - val_accuracy: 0.5783\n",
            "Epoch 27/35\n",
            "4000/4000 [==============================] - 392s 98ms/step - loss: 0.4065 - accuracy: 0.8572 - val_loss: 1.8531 - val_accuracy: 0.5675\n",
            "Epoch 28/35\n",
            "4000/4000 [==============================] - 390s 98ms/step - loss: 0.4047 - accuracy: 0.8580 - val_loss: 1.7796 - val_accuracy: 0.5711\n",
            "Epoch 29/35\n",
            "4000/4000 [==============================] - 388s 97ms/step - loss: 0.4036 - accuracy: 0.8578 - val_loss: 1.7817 - val_accuracy: 0.5707\n",
            "Epoch 30/35\n",
            "4000/4000 [==============================] - 395s 99ms/step - loss: 0.3934 - accuracy: 0.8627 - val_loss: 1.7934 - val_accuracy: 0.5759\n",
            "Epoch 31/35\n",
            "4000/4000 [==============================] - 396s 99ms/step - loss: 0.3873 - accuracy: 0.8651 - val_loss: 1.7727 - val_accuracy: 0.5749\n",
            "Epoch 32/35\n",
            "4000/4000 [==============================] - 397s 99ms/step - loss: 0.3856 - accuracy: 0.8651 - val_loss: 1.8182 - val_accuracy: 0.5724\n",
            "Epoch 33/35\n",
            "4000/4000 [==============================] - 394s 98ms/step - loss: 0.3847 - accuracy: 0.8682 - val_loss: 1.7846 - val_accuracy: 0.5683\n",
            "Epoch 34/35\n",
            "4000/4000 [==============================] - 396s 99ms/step - loss: 0.3741 - accuracy: 0.8716 - val_loss: 1.7840 - val_accuracy: 0.5673\n",
            "Epoch 35/35\n",
            "4000/4000 [==============================] - 401s 100ms/step - loss: 0.3673 - accuracy: 0.8717 - val_loss: 1.8557 - val_accuracy: 0.5707\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "def plot_history(history, title):\n",
        "    plt.figure(figsize=(10,3))\n",
        "    # Plot training & validation accuracy values\n",
        "    plt.subplot(121)\n",
        "    plt.plot(history.history['accuracy'])\n",
        "    plt.plot(history.history['val_accuracy'])\n",
        "    plt.title('Model accuracy')\n",
        "    plt.ylabel('Accuracy')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "\n",
        "    # Plot training & validation loss values\n",
        "    plt.subplot(122)\n",
        "    plt.plot(history.history['loss'])\n",
        "    plt.plot(history.history['val_loss'])\n",
        "    plt.title('Model loss')\n",
        "    plt.ylabel('Loss')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "    plt.show()"
      ],
      "metadata": {
        "id": "MvZZLKcttzkN"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plot_history(simple_cnn_model_history, 'CNN with batch size=50 epochs=15 ')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "id": "kyEE_sxbB3jv",
        "outputId": "8c11944a-134c-44fd-c46b-3b67b2cd5ab9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 720x216 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "simple_cnn_model.evaluate(test_x_reshaped,test_y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IYA5fAVKB4Gh",
        "outputId": "9fba6419-ee01-4d8f-e9d4-414ad1971d3a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 21s 65ms/step - loss: 1.8130 - accuracy: 0.5765\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.8129781484603882, 0.5764999985694885]"
            ]
          },
          "metadata": {},
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers import Conv2D, Flatten\n",
        "simple_cnn_model = Sequential()\n",
        "#---------------------------------------------------------------------------------------------------------------------s\n",
        "simple_cnn_model.add(Conv2D(64, (3,3), input_shape=(img_rows,img_cols,channels), activation='relu')) \n",
        "simple_cnn_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "# simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "# simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "# simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "# simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "\n",
        "# simple_cnn_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "# simple_cnn_model.add(Dropout(0.25))\n",
        "simple_cnn_model.add(Flatten())\n",
        "\n",
        "# simple_cnn_model.add(Dense(128, activation='relu'))\n",
        "simple_cnn_model.add(Dense(64, activation='relu'))\n",
        "simple_cnn_model.add(Dense(32, activation='relu'))\n",
        "simple_cnn_model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "\n",
        "simple_cnn_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "simple_cnn_model_history = simple_cnn_model.fit(train_x_reshaped, y_train, batch_size=100, epochs=50, validation_data=(val_x_reshaped, y_val))"
      ],
      "metadata": {
        "id": "U0TsifwdCH32",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5e3dcabc-0336-4841-8149-d34c7f56ab69"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n",
            "400/400 [==============================] - 71s 176ms/step - loss: 1.8368 - accuracy: 0.3276 - val_loss: 1.5980 - val_accuracy: 0.4303\n",
            "Epoch 2/50\n",
            "400/400 [==============================] - 70s 176ms/step - loss: 1.5239 - accuracy: 0.4559 - val_loss: 1.4552 - val_accuracy: 0.4860\n",
            "Epoch 3/50\n",
            "400/400 [==============================] - 70s 176ms/step - loss: 1.3909 - accuracy: 0.5070 - val_loss: 1.3439 - val_accuracy: 0.5180\n",
            "Epoch 4/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 1.3005 - accuracy: 0.5446 - val_loss: 1.3258 - val_accuracy: 0.5359\n",
            "Epoch 5/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 1.2476 - accuracy: 0.5579 - val_loss: 1.2465 - val_accuracy: 0.5603\n",
            "Epoch 6/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 1.1961 - accuracy: 0.5813 - val_loss: 1.2184 - val_accuracy: 0.5678\n",
            "Epoch 7/50\n",
            "400/400 [==============================] - 70s 176ms/step - loss: 1.1542 - accuracy: 0.5946 - val_loss: 1.2093 - val_accuracy: 0.5702\n",
            "Epoch 8/50\n",
            "400/400 [==============================] - 71s 176ms/step - loss: 1.1243 - accuracy: 0.6054 - val_loss: 1.1710 - val_accuracy: 0.5871\n",
            "Epoch 9/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 1.0885 - accuracy: 0.6172 - val_loss: 1.2362 - val_accuracy: 0.5691\n",
            "Epoch 10/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 1.0638 - accuracy: 0.6257 - val_loss: 1.1561 - val_accuracy: 0.5914\n",
            "Epoch 11/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 1.0353 - accuracy: 0.6358 - val_loss: 1.1595 - val_accuracy: 0.5914\n",
            "Epoch 12/50\n",
            "400/400 [==============================] - 70s 174ms/step - loss: 1.0208 - accuracy: 0.6423 - val_loss: 1.1363 - val_accuracy: 0.5983\n",
            "Epoch 13/50\n",
            "400/400 [==============================] - 70s 174ms/step - loss: 0.9934 - accuracy: 0.6529 - val_loss: 1.1368 - val_accuracy: 0.6106\n",
            "Epoch 14/50\n",
            "400/400 [==============================] - 70s 175ms/step - loss: 0.9717 - accuracy: 0.6582 - val_loss: 1.1050 - val_accuracy: 0.6216\n",
            "Epoch 15/50\n",
            "400/400 [==============================] - 70s 174ms/step - loss: 0.9577 - accuracy: 0.6646 - val_loss: 1.1077 - val_accuracy: 0.6192\n",
            "Epoch 16/50\n",
            "400/400 [==============================] - 69s 173ms/step - loss: 0.9334 - accuracy: 0.6725 - val_loss: 1.1524 - val_accuracy: 0.6044\n",
            "Epoch 17/50\n",
            "400/400 [==============================] - 69s 173ms/step - loss: 0.9184 - accuracy: 0.6775 - val_loss: 1.1381 - val_accuracy: 0.6102\n",
            "Epoch 18/50\n",
            "400/400 [==============================] - 69s 173ms/step - loss: 0.9017 - accuracy: 0.6830 - val_loss: 1.1213 - val_accuracy: 0.6176\n",
            "Epoch 19/50\n",
            "400/400 [==============================] - 69s 173ms/step - loss: 0.8878 - accuracy: 0.6880 - val_loss: 1.1098 - val_accuracy: 0.6231\n",
            "Epoch 20/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.8656 - accuracy: 0.6946 - val_loss: 1.1097 - val_accuracy: 0.6244\n",
            "Epoch 21/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.8519 - accuracy: 0.7016 - val_loss: 1.1051 - val_accuracy: 0.6226\n",
            "Epoch 22/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.8415 - accuracy: 0.7040 - val_loss: 1.1247 - val_accuracy: 0.6235\n",
            "Epoch 23/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.8227 - accuracy: 0.7117 - val_loss: 1.1289 - val_accuracy: 0.6248\n",
            "Epoch 24/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.8112 - accuracy: 0.7140 - val_loss: 1.1660 - val_accuracy: 0.6166\n",
            "Epoch 25/50\n",
            "400/400 [==============================] - 69s 174ms/step - loss: 0.7940 - accuracy: 0.7214 - val_loss: 1.1222 - val_accuracy: 0.6274\n",
            "Epoch 26/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.7821 - accuracy: 0.7237 - val_loss: 1.1336 - val_accuracy: 0.6173\n",
            "Epoch 27/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.7676 - accuracy: 0.7307 - val_loss: 1.1527 - val_accuracy: 0.6194\n",
            "Epoch 28/50\n",
            "400/400 [==============================] - 69s 171ms/step - loss: 0.7625 - accuracy: 0.7311 - val_loss: 1.1550 - val_accuracy: 0.6177\n",
            "Epoch 29/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.7488 - accuracy: 0.7350 - val_loss: 1.1552 - val_accuracy: 0.6221\n",
            "Epoch 30/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.7356 - accuracy: 0.7412 - val_loss: 1.1365 - val_accuracy: 0.6229\n",
            "Epoch 31/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.7220 - accuracy: 0.7459 - val_loss: 1.1774 - val_accuracy: 0.6167\n",
            "Epoch 32/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.7113 - accuracy: 0.7481 - val_loss: 1.1811 - val_accuracy: 0.6275\n",
            "Epoch 33/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.6981 - accuracy: 0.7521 - val_loss: 1.2442 - val_accuracy: 0.6057\n",
            "Epoch 34/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.6811 - accuracy: 0.7620 - val_loss: 1.2262 - val_accuracy: 0.6156\n",
            "Epoch 35/50\n",
            "400/400 [==============================] - 69s 171ms/step - loss: 0.6767 - accuracy: 0.7641 - val_loss: 1.2271 - val_accuracy: 0.6046\n",
            "Epoch 36/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.6659 - accuracy: 0.7654 - val_loss: 1.1907 - val_accuracy: 0.6204\n",
            "Epoch 37/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.6529 - accuracy: 0.7698 - val_loss: 1.2515 - val_accuracy: 0.6163\n",
            "Epoch 38/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.6378 - accuracy: 0.7743 - val_loss: 1.2295 - val_accuracy: 0.6172\n",
            "Epoch 39/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.6354 - accuracy: 0.7738 - val_loss: 1.2713 - val_accuracy: 0.6088\n",
            "Epoch 40/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.6187 - accuracy: 0.7829 - val_loss: 1.2762 - val_accuracy: 0.6193\n",
            "Epoch 41/50\n",
            "400/400 [==============================] - 69s 172ms/step - loss: 0.6090 - accuracy: 0.7847 - val_loss: 1.2767 - val_accuracy: 0.6179\n",
            "Epoch 42/50\n",
            "400/400 [==============================] - 69s 171ms/step - loss: 0.6008 - accuracy: 0.7863 - val_loss: 1.3144 - val_accuracy: 0.6109\n",
            "Epoch 43/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.5811 - accuracy: 0.7946 - val_loss: 1.3313 - val_accuracy: 0.5991\n",
            "Epoch 44/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.5786 - accuracy: 0.7944 - val_loss: 1.3147 - val_accuracy: 0.6128\n",
            "Epoch 45/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.5699 - accuracy: 0.7980 - val_loss: 1.3184 - val_accuracy: 0.6146\n",
            "Epoch 46/50\n",
            "400/400 [==============================] - 68s 170ms/step - loss: 0.5596 - accuracy: 0.8024 - val_loss: 1.3669 - val_accuracy: 0.6069\n",
            "Epoch 47/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.5466 - accuracy: 0.8080 - val_loss: 1.3995 - val_accuracy: 0.6045\n",
            "Epoch 48/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.5403 - accuracy: 0.8082 - val_loss: 1.4202 - val_accuracy: 0.6039\n",
            "Epoch 49/50\n",
            "400/400 [==============================] - 69s 171ms/step - loss: 0.5265 - accuracy: 0.8147 - val_loss: 1.4040 - val_accuracy: 0.6034\n",
            "Epoch 50/50\n",
            "400/400 [==============================] - 68s 171ms/step - loss: 0.5222 - accuracy: 0.8155 - val_loss: 1.5078 - val_accuracy: 0.5920\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "simple_cnn_model.evaluate(test_x_reshaped,test_y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gpH55q-RI4lA",
        "outputId": "a76a004d-b25a-405d-df3d-395d4e24a67d"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 5s 16ms/step - loss: 1.5087 - accuracy: 0.5980\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.508664608001709, 0.5979999899864197]"
            ]
          },
          "metadata": {},
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Technique 2**"
      ],
      "metadata": {
        "id": "j0JfHHudmX4a"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "num_classes = 10\n",
        "Total_train_x = np.concatenate( [train_batch_2[b'data'], train_batch_3[b'data'], train_batch_4[b'data'], train_batch_5[b'data']])\n",
        "Total_train_x = Total_train_x.astype('float32') # this is necessary for the division below\n",
        "Total_train_x /= 255\n",
        "Total_train_y = np.concatenate([np_utils.to_categorical(labels, num_classes) for labels in [ train_batch_2[b'labels'], train_batch_3[b'labels'], train_batch_4[b'labels'], train_batch_5[b'labels']]])"
      ],
      "metadata": {
        "id": "iV_PsmH0mXAH"
      },
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_val = np.concatenate([train_batch_1[b'data']])\n",
        "x_val = x_val.astype('float32')\n",
        "x_val /= 255 \n",
        "x_val.shape\n",
        "\n",
        "y_val = np_utils.to_categorical(train_batch_1[b'labels'], num_classes)\n",
        "# y_val"
      ],
      "metadata": {
        "id": "rC2eWeGlmiKy"
      },
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "test_x = test_batch[b'data'].astype('float32') / 255\n",
        "test_y = np_utils.to_categorical(test_batch[b'labels'], num_classes)"
      ],
      "metadata": {
        "id": "WAcf_s1umoZN"
      },
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_x_reshaped = x_train.reshape(len(x_train), img_rows, img_cols, channels)\n",
        "val_x_reshaped = x_val.reshape(len(x_val), img_rows, img_cols, channels)\n",
        "test_x_reshaped = test_x.reshape(len(test_x), img_rows, img_cols, channels)"
      ],
      "metadata": {
        "id": "84m1xbGQmyC1"
      },
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "img_rows = img_cols = 32\n",
        "channels = 3"
      ],
      "metadata": {
        "id": "6hZr2dPCmsvk"
      },
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers import Conv2D, Flatten\n",
        "simple_cnn_model = Sequential()\n",
        "#---------------------------------------------------------------------------------------------------------------------s\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), input_shape=(img_rows,img_cols,channels), activation='relu')) \n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "simple_cnn_model.add(Conv2D(32, (3,3), activation='relu'))\n",
        "\n",
        "simple_cnn_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "simple_cnn_model.add(Dropout(0.25))\n",
        "simple_cnn_model.add(Flatten())\n",
        " \n",
        "simple_cnn_model.add(Dense(128, activation='relu'))\n",
        "simple_cnn_model.add(Dense(10, activation='softmax'))\n",
        "\n",
        "\n",
        "simple_cnn_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "simple_cnn_model_history = simple_cnn_model.fit(train_x_reshaped, y_train, batch_size=50, epochs=25, validation_data=(val_x_reshaped, y_val))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "d26LUFmWmtZi",
        "outputId": "0d5f93c5-c88d-4419-cdc4-992c80e1b953"
      },
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/25\n",
            "800/800 [==============================] - 311s 388ms/step - loss: 1.6625 - accuracy: 0.4031 - val_loss: 1.3939 - val_accuracy: 0.4990\n",
            "Epoch 2/25\n",
            "800/800 [==============================] - 310s 387ms/step - loss: 1.3578 - accuracy: 0.5167 - val_loss: 1.2131 - val_accuracy: 0.5676\n",
            "Epoch 3/25\n",
            "800/800 [==============================] - 302s 377ms/step - loss: 1.2141 - accuracy: 0.5739 - val_loss: 1.0569 - val_accuracy: 0.6334\n",
            "Epoch 4/25\n",
            "800/800 [==============================] - 295s 368ms/step - loss: 1.1144 - accuracy: 0.6086 - val_loss: 0.9936 - val_accuracy: 0.6530\n",
            "Epoch 5/25\n",
            "800/800 [==============================] - 296s 370ms/step - loss: 1.0267 - accuracy: 0.6380 - val_loss: 0.8924 - val_accuracy: 0.6878\n",
            "Epoch 6/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.9543 - accuracy: 0.6637 - val_loss: 0.8210 - val_accuracy: 0.7155\n",
            "Epoch 7/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.8824 - accuracy: 0.6868 - val_loss: 0.7692 - val_accuracy: 0.7393\n",
            "Epoch 8/25\n",
            "800/800 [==============================] - 298s 372ms/step - loss: 0.8231 - accuracy: 0.7097 - val_loss: 0.7093 - val_accuracy: 0.7527\n",
            "Epoch 9/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.7647 - accuracy: 0.7283 - val_loss: 0.6575 - val_accuracy: 0.7796\n",
            "Epoch 10/25\n",
            "800/800 [==============================] - 297s 372ms/step - loss: 0.7116 - accuracy: 0.7484 - val_loss: 0.6005 - val_accuracy: 0.8000\n",
            "Epoch 11/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.6552 - accuracy: 0.7671 - val_loss: 0.5822 - val_accuracy: 0.8133\n",
            "Epoch 12/25\n",
            "800/800 [==============================] - 296s 371ms/step - loss: 0.6130 - accuracy: 0.7808 - val_loss: 0.5384 - val_accuracy: 0.8271\n",
            "Epoch 13/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.5715 - accuracy: 0.7967 - val_loss: 0.5197 - val_accuracy: 0.8384\n",
            "Epoch 14/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.5284 - accuracy: 0.8104 - val_loss: 0.4670 - val_accuracy: 0.8569\n",
            "Epoch 15/25\n",
            "800/800 [==============================] - 297s 371ms/step - loss: 0.4972 - accuracy: 0.8213 - val_loss: 0.4772 - val_accuracy: 0.8567\n",
            "Epoch 16/25\n",
            "800/800 [==============================] - 298s 373ms/step - loss: 0.4742 - accuracy: 0.8286 - val_loss: 0.4667 - val_accuracy: 0.8626\n",
            "Epoch 17/25\n",
            "800/800 [==============================] - 302s 377ms/step - loss: 0.4544 - accuracy: 0.8367 - val_loss: 0.4545 - val_accuracy: 0.8681\n",
            "Epoch 18/25\n",
            "800/800 [==============================] - 304s 380ms/step - loss: 0.4218 - accuracy: 0.8473 - val_loss: 0.4435 - val_accuracy: 0.8723\n",
            "Epoch 19/25\n",
            "800/800 [==============================] - 300s 375ms/step - loss: 0.4098 - accuracy: 0.8515 - val_loss: 0.4421 - val_accuracy: 0.8780\n",
            "Epoch 20/25\n",
            "800/800 [==============================] - 301s 377ms/step - loss: 0.3870 - accuracy: 0.8609 - val_loss: 0.4354 - val_accuracy: 0.8782\n",
            "Epoch 21/25\n",
            "800/800 [==============================] - 300s 375ms/step - loss: 0.3739 - accuracy: 0.8642 - val_loss: 0.4746 - val_accuracy: 0.8728\n",
            "Epoch 22/25\n",
            "800/800 [==============================] - 299s 374ms/step - loss: 0.3573 - accuracy: 0.8711 - val_loss: 0.4483 - val_accuracy: 0.8794\n",
            "Epoch 23/25\n",
            "800/800 [==============================] - 304s 380ms/step - loss: 0.3458 - accuracy: 0.8745 - val_loss: 0.4334 - val_accuracy: 0.8878\n",
            "Epoch 24/25\n",
            "800/800 [==============================] - 302s 377ms/step - loss: 0.3361 - accuracy: 0.8785 - val_loss: 0.4040 - val_accuracy: 0.8964\n",
            "Epoch 25/25\n",
            "800/800 [==============================] - 312s 390ms/step - loss: 0.3258 - accuracy: 0.8835 - val_loss: 0.4253 - val_accuracy: 0.8903\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "simple_cnn_model.evaluate(test_x_reshaped,test_y)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RT8lhPeinTz-",
        "outputId": "fea91fdb-8858-4fa7-9b15-96110171b51d"
      },
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 19s 61ms/step - loss: 0.4253 - accuracy: 0.8903\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.425336092710495, 0.8902999758720398]"
            ]
          },
          "metadata": {},
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "def plot_history(history, title):\n",
        "    plt.figure(figsize=(10,3))\n",
        "    # Plot training & validation accuracy values\n",
        "    plt.subplot(121)\n",
        "    plt.plot(history.history['accuracy'])\n",
        "    plt.plot(history.history['val_accuracy'])\n",
        "    plt.title('Model accuracy')\n",
        "    plt.ylabel('Accuracy')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "\n",
        "    # Plot training & validation loss values\n",
        "    plt.subplot(122)\n",
        "    plt.plot(history.history['loss'])\n",
        "    plt.plot(history.history['val_loss'])\n",
        "    plt.title('Model loss')\n",
        "    plt.ylabel('Loss')\n",
        "    plt.xlabel('Epoch')\n",
        "    plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "    plt.show()"
      ],
      "metadata": {
        "id": "IMdqKfhv0DwO"
      },
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plot_history(simple_cnn_model_history, 'CNN with batch size=50 epochs=15 ')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 241
        },
        "id": "klorZL6X0Ram",
        "outputId": "2f29a8c2-a90a-4337-d05b-d92198bb1aa0"
      },
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 720x216 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Time Taken** by model to train = 7528 sec = 125 mins\n",
        "# 301 sec => avg time of each epoch"
      ],
      "metadata": {
        "id": "xCH21WmBSqZ0"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "OGuxHUtn0SbC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Analysis**"
      ],
      "metadata": {
        "id": "iHh1E5YsSgEL"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Hence, it is evidenet that convolution neural network performed well with the CiFAR-10 dataet."
      ],
      "metadata": {
        "id": "5nWYWf6_Tsx8"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# And in case of time taken CNN took almost double as time taken by ANN to train the model."
      ],
      "metadata": {
        "id": "zAFxJDOAUSs3"
      }
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}